A few weeks ago I attended the first Energy Research and Social Science conference in Sitges near Barcelona. I learned lots of things about social scientific approaches to studying energy. There were some particularly good talks on fuel poverty research – the new European energy poverty observatory is well worth checking out – and I was also introduced to the field of energy justice. The talk I gave wasn’t directly related to fuel poverty, but rather was the result of work the research group I’m in has been undertaking to develop a toolkit for energy research which aims to help avoid some of the potential pitfalls of the research process.
With this in mind, I thought it might be good to outline that process here. This is most relevant to quantitative research, so it doesn’t really cover qualitative methods. This isn’t an exhaustive or particularly critical overview, but rather an attempt to distil the process into a simplified arc. It’s definitely idealised, more often than not the actual practice of research is littered with uncertainties and is contingent on external forces, such a funding streams, changing time frames and interactions with colleagues.
0: Decide what to look at. Ask yourself why you want to know about something, or more crucially, why you want to invest time and effort understanding something rigorously. Often interest in a field can be motivated by personal experience, which is sometimes beneficial, but can bring preconceptions. Equally, wading into an area you have absolutely no knowledge of can be challenging. The general rule of thumb is that any idea you feel is new probably isn’t. Practically there are constraints on this depending on your expertise. Tools like Free-Mind can be useful at this stage.
1: The literature review. A review of existing thought on a subject allows you to identify areas about which little is known. This stage is essential for forming meaningful research questions or study designs. There are excellent tools available for organising literature. I personally use Zotero, and have collected a few thousand references in my personal library over the years, not all of which I’ve read in detail I should admit. Recently, interest has developed in the idea of the comprehensive literature review, which are much more systematic than previous approaches. Search engines like Scopus and Google Scholar are indispensable.
2: Formulate research question and methodology. This stage involves formulating questions which could address the gaps or problems which you identify from your literature review. The bundle of methods you use to answer these questions is your methodology. The methodology you chose will usually be related to the discipline you come from, and some topics are more suited to certain questions than others. Ideally, in the quantitative sciences at least, you will come up with hypotheses at this stage, which you record and don’t adjust once you’ve seen the data.
3: Data collection. Here the term data doesn’t need to be numbers or measurements per se, but could be collections of opinions elicited from interviews or information from other sources. The validity and stability of your data depends again on your methodology. Some methodologies, especially within linguistics or other language based fields, produce results which depend on the actions of the investigator. This doesn’t make them less useful than numerical based approaches, it just means the research is doing a different thing than numbers do. Some of the most problematic research results come from the assumption that numerical methods can be imposed over systems which don’t have an underlying numerical structure.
4: Analyse and interpret. Under best practice you will have pre-specified your analysis techniques before collecting data, particularly in situations where statistics are used. There are thousands of statistical tests that can be used to slice data almost any which way you can imagine, so making sure they relate to the research question before data collection is vital. Accurate uncertainty analysis is one of the hardest and most important things to do. I will save thoughts on Bayesian approaches for a future post.
5: Report, publish and share. This is the stage which links your work back to the literature. Assuming you’ve met the criteria that a particular field sets for considering something worthy of publication you might submit it to a journal. More informal pieces like this probably belong in blogs, where almost anything goes. Your work might even stimulate a paradigm shift in understanding, but these are extremely rare, and this might not actually be how the history of science progresses anyway.
For the sake of approximate completeness, there is Feyerbrand’s Against Method, which I haven’t read closely, but I mention because I don’t want to give the impression that I subscribe to a single all encompassing method of finding things out. While I don’t agree with Feyerband, I think we’re at a particular moment in history where certain anti-scientific political movements have produced a somewhat alienating pro-science reverie which is problematic in a different way, and so a little caution is required.